How to analyze distortion

[lowell]

I'm curious on how everyone might analyze distortion. Whether it be on a scope, analyzer etc... I've been scoping a custom amp and am not clear on how to decipher some clipping that I'm seeing. Any input appreciated, whether it be about reading scope waves or using another device like an analyzer.

[R.G.]

I'm curious on how everyone might analyze distortion. Whether it be on a scope, analyzer etc... I've been scoping a custom amp and am not clear on how to decipher some clipping that I'm seeing. Any input appreciated, whether it be about reading scope waves or using another device like an analyzer.

If you're seeing clipping on a scope, the tool to use is a scope. If you're seeing faint bends in the waveform, scopes are too clumsy to use and you need one of the probably 87 kinds of distortion analyzers.

Be aware that clipped waveforms run through any kind of filtering will have their components phase shifted and won't look simply clipped any longer.

[Enzo]

There is a whole lot of measurable distortion that you won't easily see on a scope. CLipping is obvious to the eye, but if a sine wave is a litle rolled off on the tops - in other words a little fatter than it should be - it may appear perfectly normal to the viewer.

I have a distortion analyzer here. The only thing I ever use it for is once in a while as an AC voltmeter. I was required to have one as a QSC service station, but my ears tell me a Fender sounds like a Fender, not the distortion meter.

[lowell]

Ok so if I'm building custom tube amps and want to know why i have a wacky looking wave, like a sawtooth with scooped peaks and valleys, will a distortion analyzer decipher phase shift and harmonic content? I guess according to Enzo you don't find it useful for the work you do, but are you tweaking tube distortion circuits? Just curious if I should spend the money or if there's another way to go about this.

[Enzo]

It is an opinion only, I do maintenenace and repair. There are many tools that different techs have different opinions about. Like those little things for your scope that are crude curve tracers - the Huntron Tracker is an overpriced example of one made commercial. You hook it to your scope and put a part across its leads. A resistor makes a straight sloped line a cap makes a loop, a diode makes a flat line with a 90 degree bend. A quick check. I never found it useful, but some techs use it every day and would hate to lose it. I find no use for a distortion meter - I'll make my case below. Someone else may love it.

I look at circuits like I look at them - everyone does it different. I deal in diagnostics. To me a distortion meter is less diagnostic and more assessment related. Say you expected a nice clean waveform in a circuit, you connect the distortometer, and sure enough, 78% distortion. Now what do you know about its source that you didn;t already? And if there were a phase meter, what would it tell you?

Of course if I am making hifi circuits and want to minimize distortion and maximize fidelity, then certainly it becomes useful.

You might find more useful - or not - some analytical software that displays frequency versus amplitude versus phase or time or whatever. There are a number of audio programs that do this - I got no names at the moment. The graphic on the screen looks a bit like a 3-D mountain landscape moving along. That might reveal something about freq response etc.

If you are analyzing distortion circuits, for example, from a particular point in development, would you want just more distortion, or maybe a different sort of distortion? Just like some guitar amp where it really sounds better with the gain on 6 and the master up a little than the gain on 10 and the master down. Tweaking a circuit component might increase distortion 5% on the meter, but is that useful? MAybe, but to me that lacks context. 25% of this sort of distortion won;t sound like 25% of that kind of distortion, but they both will say 25% on the meter.

DO you know how distortion analyzers work? You take a test signal and feed it to both the amp or circuit under test, and to the refernce input of the analyzer. Then you also run the output of the circuit uner test to the other input of the analyzer. One is in reverse phase to the other, so the signals cancel out. We adjust for levels, and when the dust settles, the input signal and output signal out of phase combine and cancel EXCEPT for that part of the output that wasn;t part of the input. SO if the input and output were EXACTLY the same, then when the levels were equal, they would exactly cancel and the result would be nothing.

When they are not identical waveforms, then the difference is by definition distortion. We then measure the amount of difference to quantify that distortion.

Now more modern fancier units may have additional features, but that is the basic idea.


If you are having clipping in a circuit that you don;t want, a distortometer will tell you that it is there, but it won;t tell you why?


But I don;t want to shoo you away from something you might find useful. Maybe there is some place or someone that you could borrow one from to find out?

[Steve Conner]

I use a scope for "eyeball" work, and for finer analysis, I'll use audio spectrum analyzer software running on a PC with a 24-bit pro soundcard. This works considerably better than those old distortion meters did, though it's not up to the standard of say an Audio Precision test set.

My current favourite is Spectrum Lab, which does the "3-D mountain thing" fine, but Rightmark Audio Analyzer seems good too.

[MWJB]

"Distortion" is a phrase that covers a wide spectrum of sounds, some pleasing to the ear, some not. As Enzo alludes, detecting/analysing it might not tell you what category it falls into. If you are scoping lots of known quanities, then no doubt you can build a reference that will show when something might be distorting abnormally. But, if as you say, you are building custom (one off, or small bespoke runs) guitar amps, then you are somewhat out on a limb in that respect...the only way to evaluate them is to listen to them. You might find that, for instance, in the case of some SE amps, that in the territory where they are typically played, they are constantly distorted, yet whether they sound good or not will depend on what that particular distortion sounds like.

If you are concerned about going deaf/inciting the neighbours to lynch you, or other problems associated with playing amps at driven volumes, perhaps invest in a speaker simulater that you can run out to a mixer/recording device. Obviously the speaker you go with at the end of the day will need appraising, but this might help you get some of the way.

Eventually experience and familiarity with the circuit will allow you to analyse which part of the circuit requires attention & never discount the usefulness of taking dc measurements too (typically I have always been able to find the source of a circuit malfunction by checking dc), these often indicate symptoms that might be behind unwanted, or excessive distortion.

Remember that an amp is a musical instrument, it's sole purpose is to deliver a pleasing sound to human ears, makes sense to use those ears to evaluate it.

[lowell]

Ok so maybe there's an easier way to get the answers I'm looking for since all those software/tools seem a bit over the top for my uses. Does someone know of a tutorial on reading waves on a scope? IE if I have a square wave w/ scooped peaks and valleys what does that tell me about frequency, phase etc..??

[Old Tele man]

...in the old days it was done directly off the control grid (EgIb) and plate (EbIb) curves using only simple mathematical equations and selected reference points from both sets of curves.

...see RDH3 (or 4) for details.

[olddawg]

Ok so maybe there's an easier way to get the answers I'm looking for since all those software/tools seem a bit over the top for my uses. Does someone know of a tutorial on reading waves on a scope? IE if I have a square wave w/ scooped peaks and valleys what does that tell me about frequency, phase etc..??

If you have a decent dual channel scope, a simple way to start is to reference a 1K tone on the input and observe the output. Make the 2 waveforms the same size on the scope and display one cycle. Now superimpose one over the other. The difference will be some of the distortion in the output. You will probably trigger off of the source. Your scope may also give you the option to invert one one channel or the other and combine the A and B channels. If it is a good scope you can expand it out to look at specific parts of the waveform. If you do that you may be able to see the differences that Enzo is talking about. Of course you are only looking at a simple tone at a single frequency and there are lots of other factors. But it's a start.

[lowell]

ok thanks everyone. I'll try some of those techniques and see if I can learn more about what I'm hearing/seeing on the scope.

[MWJB]

It might also help to put what you are seeing in context - what sort of clean power are you getting before distortion sets in (is it on par with similar output topologies), does it appear earlier in the circuit first? Schem, preferably with dc & AC voltages?

At the moment you seem to be saying, "I have a one-off amp with a funny looking wave, why?", but without giving any clues. You may as well be asking why your big toe hurts...you could get 100 well meaning, but irrelevant, responses before you mention the grand piano resting on it.

[lowell]

here's the schem. At the grid of V4 I'm getting clipping on positive cycle of wave, not negative. I understand that this is grid blocking distortion due the the input voltage exceeding Vg/k. I know decreasing coupling caps, increasing grid stoppers, decreasing bypass caps etc... reduces it. I'm curious as to why the length of the clipped part of the wave is twice as long as the width of the unclipped negative portion of the wave at the same points respectively:: .6v and -.6v on the scope. Sorry I posted bout this in another thread too.

I'm not concerned bout WHY I have distortion or how to get rid of it, because obviously this is the OD channel. more curious how to interpret what my eyes see on the scope.

[MWJB]

Or V1b is under biased (what happens when you double R12?). There are no plate voltages listed either?

Either way you have little attenuation between V1b & V4a. With that plate fed tone stack following yet another triode (assuming that V4 is one triode per stage, not parallelled triodes as drawn) & volume control, you have an awful lot of gain there. I would move the volume pot to after the first stage (you still will most likely need to tweak pot value with another resistor tacked between wiper & ground) & lose a triode if you want to stick with a plate fed tone stack. However, as you have the extra triode to hand, why not convert V4 to a cathode follower stage?

[lowell]

Voltage on OD stage plates are around 250v. Oh and C6 is actually 1uf, R41 is 100k, R18 is 1m.

V4 IS paralleled triodes. Are you suggesting no gain at V4 and just turn it into cathode follower altogether? If so I'll lose gain AFTER the OD stage right? Not sure I want to do that. Do you think there's too much gain reaching V2A? Just curious.